Study Of Remote Sensing Image Fusion Algorithm Based On Multiscale And Sparse Representation

Based on the environmental monitoring of industrial solid waste in Ningxia,this paper studies the fusion algorithm of high resolution panchromatic image and multispectral image for the characteristics of high-resolution imagery of industrial solid waste.The fusion of high resolution panchromatic image and multispectral image enables the fused image to have both high spatial resolution of panchromatic image and rich spectral resolution of multispectral image.In this way,the difference and limitation of single sensor in terms of spectral and spatial resolution can be overcome.Firstly,this paper studies the fusion algorithm based on the combination of dual tree complex wavelet transform and sparse representation.This algorithm uses the dual tree complex wavelet method to decompose the high resolution panchromatic image and multispectral image,and then obtains the high frequency subband and the low frequency subband.Then,the low-frequency subband is extracted detailed information and fused by sparse representation method,and the high-frequency subband is fused by the fusion strategy combining large local energy and weighted average.Finally,the fusion image is reconstructed by the dual tree complex wavelet inverse transform of the fused subband.The experimental results show that this algorithm has a better image fusion effect than other algorithms such as dual tree complex wavelet transform,sparse representation and curve transform.Secondly,the paper studies the fusion algorithm based on the combination of shearlet transform and sparse representation.This algorithm uses the shearlet transform method to decompose the high resolution panchromatic image and multispectral image.After decomposition,the high frequency coefficients and low frequency coefficients are obtained.Then,average fusion was performed on the low-frequency coefficients,and the over-complete dictionary was obtained by combining with K_SVD algorithm.The absolute value of fusion coefficients was calculated,and the fusion coefficients with larger absolutevalue were selected correspondingly to obtain the new low-frequency sub-band coefficients.SML-based fusion rules are used to fuse the high frequency coefficients.Finally,the shearlet inverse transform is used to process the low frequency and high frequency coefficients to get the final fusion image.The nonsubsampled shearlet transform has the characteristics of multiscale,multi-direction and translation invariant,which can make the fused image retain more details in the source image and effectively eliminate frequency aliasing.The experimental results show that the two fusion methods are better than other traditional fusion methods in image definition,spectral information retention and noise control.On the basis of the above fusion algorithm,a remote sensing image fusion system with visual interface is implemented in this paper,which provides a friendly interface for the study of the fusion algorithm in this paper.